Exploring Spatial Trends and Influencing Factors for Gastric Cancer Based on Bayesian Statistics: A Case Study of Shanxi, China

Gehong Zhang, Junming Li, Sijin Li, Yang WangInternational Journal of Environmental Research and Public Health 2018 · DOI

TL;DR

This study tackles the challenge of mapping gastric cancer (GC) risk and identifying its drivers in Shanxi, China, using hospital data. By developing a novel Bayesian spatial statistical model that corrects for biases in hospital-diagnosed cases, they reveal a distinct "high in the south, low in the north" spatial pattern of GC risk. Their analysis points to dietary habits and geographic factors as key influencers, offering valuable insights for targeted public health interventions.

What Problem Does This Paper Address?

Gastric cancer (GC) is a major global health concern, particularly in East Asia, and China bears a significant burden. Understanding the spatial distribution of GC and the factors contributing to it is crucial for developing effective prevention and control strategies. However, obtaining reliable epidemiological data for disease mapping is challenging due to the labor-intensive nature of surveys and logistical constraints. While hospital-diagnosed data is more accessible, it can be biased, especially when collected from a single institution. This paper aims to address this by developing a statistical framework to map GC risk using hospital data and identify its influencing factors, specifically in Shanxi Province, a region known for its high GC incidence.

Methods

The core of this research lies in a two-pronged statistical approach:

  1. Bayesian Spatial Statistical Model with a Selection Probability Model: To overcome the bias inherent in using hospital-diagnosed data from a single hospital (the First Hospital of Shanxi Medical University), the authors developed a "selection probability model." This model estimates the probability of patients from different regions choosing that specific hospital, allowing them to correct the raw case counts for each city. This corrected data is then fed into a Bayesian spatial statistical model (specifically, a BYM model) to estimate the spatial trends of GC risk. This Bayesian approach is well-suited for handling spatial dependencies and uncertainty, especially with limited data.
  2. Bayesian Lasso Regression Model: To identify the factors influencing GC risk, the researchers employed a Bayesian Lasso regression model. This technique is particularly useful when dealing with a large number of potential predictor variables (22 in this case) and a relatively small sample size, as it performs variable selection and shrinkage simultaneously, helping to mitigate multicollinearity and overfitting.

The study normalized all variables by dividing by their provincial average to account for dimensional effects before applying the models. Model convergence was assessed using the Gelman-Rubin statistic.

Datasets

The primary dataset used in this study consists of 358 gastric cancer (GC) patients diagnosed and treated at the First Hospital of Shanxi Medical University (FHSMU) in Taiyuan, Shanxi Province, between April 2014 and June 2016. These patients resided in 11 prefecture-level cities within Shanxi.

In addition to patient data, the study utilized various socio-economic, dietary structure, medical condition, and geographic environment variables for the 11 cities.

  • Socio-economic and dietary data were primarily sourced from the Shanxi Statistical Yearbook of 2015.
  • Geographic environment data (temperature, terrain, vegetation index, precipitation) were obtained from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC).

No new benchmark datasets were introduced; the focus was on applying advanced statistical methods to existing data sources.

Key Insights and Ideas

  • Hospital Data Can Be Mapped: The development of the "selection probability model" is a significant contribution, demonstrating a viable method to correct for selection bias in hospital-diagnosed data, making it useful for spatial disease mapping where survey data is scarce.
  • Shanxi's GC Hotspot: The clear spatial pattern of GC risk ("high in the south, low in the north") is a critical finding, suggesting that regional factors play a substantial role and that interventions might need to be geographically targeted.
  • Dietary Diversity Matters: The finding that a more diverse dietary structure (associated with less reliance on grain except corn/wheat, and less beef/cow milk production) is linked to lower GC risk is a nuanced insight, potentially highlighting the protective effects of varied nutrient intake.
  • Geographic Factors are Crucial: The strong association of temperature, terrain, vegetation, and precipitation with GC risk underscores the importance of environmental determinants in disease patterns.

Key Results

  • Spatial Trends: The study identified a distinct spatial pattern of GC risk in Shanxi, characterized by higher risk in the southern regions (particularly Jincheng and Changzhi) and lower risk in the northern areas. The highest GC relative risk observed was 1.291 (95% HPD: 0.789–4.002) in Jincheng.
  • Influencing Factors (Univariable & Bayesian Lasso):
    • Negative Associations (Lower GC Risk): Increased consumption of beef and cow milk, and a larger sown area of grain except for corn and wheat, were significantly associated with lower GC risk (p ≤ 0.08).
  • Positive Associations (Higher GC Risk): Pork production per capita was positively correlated with GC risk.
  • Geographic Factors: Temperature, terrain, vegetation cover, and precipitation all showed significant associations with GC risk (p < 0.05), with higher temperatures and precipitation linked to higher risk.
  • Socio-economic Factors: A higher proportion of living expenditures to disposable income per capita of urban households (indicating less savings/prosperity) was associated with higher GC risk (P(β4>0|y) = 82%). Conversely, a higher percentage of tertiary industry (indicating a more developed economy) was associated with lower GC risk (P(β3<0|y) = 71%).

Limitations and Future Work

The authors acknowledge several limitations:

  • Small Sample Size: The study was based on data from a single hospital, resulting in a relatively small sample size for the 358 patients. This limits the precision of the estimates.
  • Data Availability: Direct survey data for the same period was unavailable for verification, necessitating reliance on previous studies for validation.
  • Unexplored Factors: Factors such as regional salt consumption and vegetable/fruit production were not included due to data limitations but are suggested for future research.

Future work could focus on:

  • Larger Sample Sizes: Including more patient data from multiple hospitals or conducting broader epidemiological surveys would enhance the precision and generalizability of the findings.
  • Spatio-temporal Trends: Investigating how GC risk patterns evolve over time would provide a more dynamic understanding of the disease.
  • Additional Factors: Incorporating other potential risk factors like salt intake and specific dietary components could refine the analysis.
  • Mechanism Exploration: Further research is needed to elucidate the precise biological and environmental mechanisms underlying the observed associations.

Summary updated

Abstract

Gastric cancer (GC) is the fourth most common type of cancer and the second leading cause of cancer-related deaths worldwide. To detect the spatial trends of GC risk based on hospital-diagnosed patients, this study presented a selection probability model and integrated it into the Bayesian spatial statistical model. Then, the spatial pattern of GC risk in Shanxi Province in north central China was estimated. In addition, factors influencing GC were investigated mainly using the Bayesian Lasso model. The spatial variability of GC risk in Shanxi has the conspicuous feature of being 'high in the south and low in the north'. The highest GC relative risk was 1.291 (95% highest posterior density: 0.789⁻4.002). The univariable analysis and Bayesian Lasso regression results showed that a diverse dietary structure and increased consumption of beef and cow milk were significantly (<i>p</i> ≤ 0.08) and in high probability (greater than 68%) negatively associated with GC risk. Pork production per capita has a positive correlation with GC risk. Moreover, four geographic factors, namely, temperature, terrain, vegetation cover, and precipitation, showed significant (<i>p</i> < 0.05) associations with GC risk based on univariable analysis, and associated with GC risks in high probability (greater than 60%) inferred from Bayesian Lasso regression model.
BibTeX
@article{zhang2018exploring,
  author = {Gehong Zhang and Junming Li and Sijin Li and Yang Wang},
  title = {{Exploring Spatial Trends and Influencing Factors for Gastric Cancer Based on Bayesian Statistics: A Case Study of Shanxi, China}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2018},
  doi = {10.3390/ijerph15091824},
}
Figure